CycleLens Open the dashboard

Methodology, sources and corrections

CycleLens shows a lot of numbers. This page explains where they come from, how they are dated, what a model reading is and is not, and what happens when something is wrong.

Three kinds of statement

Everything the product says is one of three things, and it is labelled as such:

  • An observation — a figure a source reported, with its time: a price, a market value, a funding rate.
  • A model reading — what one of the product's named models computes from those figures, with its horizon: a confluence score, a position in a fitted channel, a calendar phase.
  • An interpretation — what the observations and readings together are consistent with, and what they cannot establish.

Data sources

Each figure names its source beside it. These are the sources the product reads, and what it reads from each:

Data sources and what each supplies
SourceWhat the product reads from it
CoinGeckoprices and daily price history; market value and the global total; Bitcoin and Ether share of that total; reported 24-hour volume; market-value ranks; category tags; code-activity and audience counts; trending searches; project descriptions and links; circulating, total and maximum supply; open interest and funding on the futures exchanges in its derivatives list; each exchange market’s spread and cost to move the price 2%; a pool’s candles and large trades, and a token’s largest holders (its on-chain data)
alternative.methe daily Fear & Greed index
Glassnodeon-chain series — NUPL, SOPR, MVRV, exchange flows, active addresses
Whale Alertlarge exchange transfers above a size threshold
each publisher’s own pagethe publication time of an article a news search returned
Hyperliquidits perpetual markets each minute: open interest, funding, the futures and index prices and the cost of a set trade; every public trade, totalled by buyers and sellers each minute, with the addresses of the largest takers; the order books of its fifty largest markets each minute; its one-minute, five-minute, hourly and daily candles; the hourly funding history of its perpetual markets; the open positions of the few largest takers a report names

Powered by CoinGecko. Some panels need an access key for their source; where one is not set, the panel says so rather than showing an empty figure.

Time and freshness

  • Days are UTC days. A daily close belongs to the UTC day it closes; a day still in progress is marked as running and is never treated as a close.
  • Every figure carries its time. When a source is slow, the latest saved copy is shown with the time it was taken, never passed off as current.
  • Publication time is not event time. A news story is placed by when it was published, and that is compared with when the move it discusses began.

Models and their limits

Each model is named with what it reads and over what horizon. Where a model has not been tested against what followed its readings, the product says so; no reading is presented as a forecast, a signal or a recommendation. A reading on a coin you hold describes the coin — it is never an instruction about your position.

Forensics reports

A Forensics report measures a move on five-minute bars over the last 24 hours and times its first break on one-minute bars. It records the order in which things happened — the first unusual move on each market, the pick-up in selling, the fall or rise in open interest — and keeps that order apart from cause. It reads the coin's perpetual market on Hyperliquid, its own trading pools and its market price across exchanges, and does not show forced closures or the long/short ratio, which no source it may show publishes. It never names who traded or why. Every report lists what it established and what it did not, and which markets and feeds it could not read.

How AI is used

  • The assistant answers questions about what is on the screen using Anthropic's Claude. It can be wrong, and each answer names the figures it used.
  • Forensics reports are measured by the product's own code. A language model writes a few paragraphs on top, and those are checked against the measured figures before they are shown: a sentence that quotes a figure the report did not measure, gives advice or claims a motive is refused.
  • Articles on this site are drafted with AI assistance and are published only after the founder has fact-checked each one against its sources and approved it; each says when.

Editorial policy

  • Nothing on this site is paid for by a project, an exchange or an advertiser, and there are no affiliate links.
  • Every article carries the same standing disclosure: the founder may hold or trade crypto assets, including ones an article discusses, at any time.
  • Examples use public data or clearly labelled sample figures — never a reader's own portfolio or report.
  • An article shows when it was published and when it last changed meaningfully. A typo fix does not move the date.

Corrections

If you find an error, write to support@cyclelens.ai with the page and the figure. A confirmed error in an article is fixed and a dated correction note is added to the page. A Forensics report that was saved before a correction to the method keeps what it said when it was filed, and opens under a notice listing what has been corrected since — reports are never rewritten silently.

Questions about a figure?

Figures name their source beside them; for anything else, write to support.

Contact support

Open the dashboard